A new scheme to reduce leakage in deep-submicron cache memories with no extra dynamic consumption